Re: CHECKPOINTS QUESTION

From: varad acharya <vacharya.ford_at_e-mail.com>
Date: 1996/10/08
Message-ID: <325A3F95.1EF8_at_e-mail.com>#1/1


David J Roth wrote:
>
> Alexander Bardos wrote:
> >
> > Hi,
> >
> >
> > My question:
> >
> > What happens when an instance runs out of db buffer cache blocks (
> > unmodified ) and there are only dirty blocks available? Is this possible
> > at all? Shouldn't the DBWR flush the dirty blocks ( write to OS files /
> > partitions -> do a checkpoint )?
> > Are the dirty blocks managed on separate lists that are not subjected to
> > the LRU realocation algorithm?
> >
> > Thanks.
> >
> > --
> > ------
> > Alex Bardos <alexbrds_at_unix.infoserve.net>
> > ------

The DBWR process will write dirty buffers to disk under these conditions

A user process when moving a buffer to the dirty list finds that the dirty list has reached a threshold value will signal the DBWR to write dirty buffers.

When a user process scans the LRU list beyond a threshold value and finds no free buffers will stop searching and signal DBWR to write.

When a timeout occurs (3 sec)

When a checkpoint occurs.

Refer the Oracle Server Concepts manual (7.2) for more info on the Oracle background processes.

-- 

Regards,

Varad Acharya
Received on Tue Oct 08 1996 - 00:00:00 CEST

Original text of this message